Investigate replacing MethodRun.predictive with stagereg PosteriorStage.predict
Nobody has claimed this yet.
Assessment
- Difficulty
- 4/5
- Estimated time
- 1-2 days
- Newbie friendliness
- 48/100
- Issue type
- Refactor
- Clarity
- Mostly clear
- Activity status
- Active
- Tech stack
- jupyter-notebook
- Domain
- backend
Research direction
Start by reading MethodRun.predictive and the restore flow where MethodRun is populated; the issue names these entry points but gives no file paths. Compare its predictive with PosteriorStage.predict(inputs, key) from stagereg 0.3.0, including the aleatoric-variance mean, and check how HNN's path would fit. Evaluate test-set performance, especially for pSGLD; done means the equivalence, restore-flow impact, split-path simplicity, and speed tradeoffs are documented.
Written by the indexing model from the issue text.
Description
Question
stagereg 0.3.0 adds PosteriorStage.predict(inputs, key): a compiled PosteriorPredictive after train or restore. qmodem builds its own predictive in MethodRun.predictive, using the same aleatoric-variance mean.
Can qmodem delegate to PosteriorStage.predict?
To investigate
- Does it give the same predictive as qmodem's, up to floating point?
- HNN has no posterior stage, so it still needs qmodem code. Does a split path stay simpler than today's?
MethodRunwould have to hold the stage after restore. What does that change in the restore flow?- Is it faster on test-set evaluation, especially for pSGLD?
Out of scope of the stagereg 0.3.0 bump; recorded during that grilling session.
- Dominant language
- Jupyter Notebook
- Stars
- 0
- Forks
- 0
- Avg merge
- 1d 11h
- Merged PRs (30d)
- 7
Getting set up
- No Dockerfile or Docker Compose file
- No pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from GroupiSP/qmodem
-
enhancement needs-triage
Difficulty 4/5 3-5 days Newbie friendliness 20/100
-
needs-triage
Difficulty 5/5 Over a week Newbie friendliness 35/100
-
enhancement
Difficulty 4/5 3-5 days Newbie friendliness 15/100
-
Consistency: specify a global precision for all numpy/JAX arraysMay be free again @debrevitatevitae claimed this 88 days ago, and no pull request is open. Openreproducibility
-
CNN: replace pooling with a flattening layerMay be free again @debrevitatevitae claimed this 88 days ago, and no pull request is open. Openenhancement
Similar issues
-
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
-
Discovered ports from service discovery are never propagated to modules in Module.load()Possibly taken @vishalmore90 claimed this today. Open
Difficulty 2/5 1-3 hours Newbie friendliness 70/100
OWASP/Nettacker#1828 · 1 comment ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 62/100
voidzero-dev/void#677 ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 80/100
sjkim1127/Reversecore_MCP#363 ·
Maintainers usually reply within 1 day
-
bug
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
lidge-jun/opencodex#8101 · 2 comments ·
Maintainers usually reply within 1 day